A Research On Liquidity Risk Based On The D-vine Copula

Liquidity means the ability of assets' quickly realization. On the stock market, market liquidity risk refers to economic losses uncertainty which caused by the generate transaction difficulties or increasing of transaction costs when the market lacks liquidity. Financial market volatility is constantly increasing in recent years, especially in the 2008 US sub-prime mortgage crisis, which swept most of the world economies triggered a global financial tsunami. Liquidity risk is an important factor which evolved this sub-prime credit crisis into a global financial crisis.The liquidity pressures were transferred out when the investors change their money to other markets for the sake of obtaining liquidity and then Liquidity risk in the case of market panic spread among different markets and financial systems.From mid-June 2015 to early July, the drop of China's stock market is even up to 30% after falling through three consecutive weeks, and that made about 20 trillion RMB evaporated. Stock market crash triggered panic and stampede, the venue of drying up of liquidity even fast shocks to other financial markets. This liquidity crisis was not resolved until regulators promptly bailed it out and these traumatic experience confirms liquidity risk is one of the major financial market risks. Financial markets and institutions will suffer huge losses if they are short of liquidity risk management or ignore liquidity risks.How to accurately measure the liquidity risk and its effective financial management is the direction which has been actively explored by our financial industry. The liquidity risk-adjusted VaR (Liquidity adjusted VaR, La-VaR) is a risk factors of integrated risk measurement methods which comprehensive considered the market and liquidity risk, and the liquidity indicators can be divided into the spread, holding time and volume. However, for portfolio risk management, the general La-VaR is still not perfect because it neither pays any attention on the related structures between market risk liquidity risk, nor measures the dependencies of assets accurately Appearance of Copula theory provides a good solution for above problems. The Vine Copula has been vigorously promoted by both of the theorists and practitioners because it can be more flexible to reflect the related structures between different variables, especially the tail dependence. In China, Vine Copula has been widely used in empirical finance like Wuyi Ye et al. (2015) used vine Copula to estimate the self-dependence structure in trading volume duration and he got a good prediction result.Jiang Gao(2013) selected Pair-Copula module with characteristics of tail of the distribution to construct vine Copula model predicting the portfolio VaR. Ziping Du and Xuefeng Zhang (2014) concluded that foreign exchange portfolio based on the mix of Vine-C can simulate better forecast by comparing with other VaR methods. But Vine Copula research based on La-VaR was not yet studied.Summarize the shortcomings of the current La-VaR study is that it's insufficient to portray the dependencies between market risk factor and liquidity risk factor.For the La-VaR portfolio, the correlation between asset returns of each measure is relatively simple. As we all know the market risk and liquidity risk have thick tails, accurately grasp the tail dependencies between risk factors will help to improve the predictive accuracy of La-VaR. So, which kind of liquidity indicator is more appropriate,how to choose a liquidity risk model which can be embedded vine Copula methods, and how to design with a realistic and workable framework Copula vine is the main problem to be solved herein.This paper learns from the mature theories and empirical experience, choose a liquidity risk measurement models from the research frontier, and it must meet the following three conditions:(1)No directly applies from the foreign quote market mechanism model, but applies to suit our domestic instruction-driven market mechanisms;(2)Not only single asset measurement, but also more effective measures to integrate liquidity risk by adjusting the portfolio;(3)Model method structure with the use of Vine Copula.Subsequently, select risk factors and risk factors Copula embedded way according to the liquidity risk framework. Considering the high frequency data can capture more market volatility information, this paper tries to use a high-frequency tick data from stock combination to go empirical research. By calculating the value of La-VaR based on Vine Copula, finally make the appropriate test to validate the model results.This paper combines theoretical and empirical research methods. In theory, this article improve the classical BDSS framework based on the bid-ask spread liquidity risk measurement methods. In practical, I select the optimal fitting model based on empirical characteristics of high frequency trading spreads and feature of profit rate, then measure the results of the portfolio La-VaR calculation depends on Copula. Secondly, this article makes full use of time series method and econometric model to analyze and modeling, and capture the sharp peak and heavy tail as well as heteroskedasticity characteristics of profit rate series in real market by constructing t-GARCH model. And use autoregressive conditional dual Poisson model to model the absolute spread sequence. This article also applies to comparative analysis. Try to use absolute spread instead of relative spread on the choice of Liquidity risk factor, and compare the effect of La-VaR forecast under two different indicators to judge whether using absolute spreads as liquidity risk factor under the stock market environment is reasonable. Compared with the traditional VaR, it inquiries the liquidity risk characteristics of China's stock market.The full text is a total of six chapters. The first chapter is an introduction, leading the research background, based on the theoretical and practical background of this study to extract the theoretical and practical significance and elaborating the main contents of this paper, the methods and innovation one by one at the same time. The second chapter is literature review which is divided into four sections. First section is the liquidity overview, followed by arranging of liquidity risk measurement models, and then showing the latest research results in the field of vine Copula, finally summarizing for current literature. The third chapter is the theoretical method, selecting a re-amended BDSS model adapted to domestic securities market mechanism according to the theoretical study of scholars" mobility in the field, with introducing the model which is under the framework of market risk factors returns and liquidity risk factor quote spread (including absolute and relative spread index). Then this article introduces Vine Copula model and parameter estimation methods briefly.Finally the selection of data will be described. This paper selects three stocks' five-minute high-frequency tick data from the Chinese A-share stock market for 21 trading days of November 2015.The fourth chapter is the empirical research.We take model fitting and parameter estimation on return rate, relative spread and absolute spread sequence. Next estimates Copula coefficient among every stock risk factor on the premise of relative spread index to obtain La-VaR, and then re-estimate Copula coefficient to obtained La-VaR'by using the absolute spread index as liquidity risk factor, determine whether the market liquidity risk exists or not by comparing with historical simulation based on VaR values of traditional method.Finally verify the validity of the model though Kupiec failure rate test to determine the merits of the spread index and summarize the reasons. The fifth chapter is the conclusion, which also summarize the empirical results and make recommendations based on current situation of the stock market, at last of article deficiencies and Prospect are discussed.The empirical results show that the rate of returning stocks from different industries have a certain degree of symmetry tail dependencies, and between the yield of each stock and the bid-ask spread, there are some tail dependencies too. Portfolio risk measurement between these risk factors in different complex dependence structure can not be ignored, and abundant Pair Copula function family have provided favorable conditions to accurately capture the correlation between the composition of assets internal differences. In three confidence levels, D-Vine Copula methods based on the relative spread La-VaR and absolute spread La-VaR'both get higher values than the historical simulation method VaR indicating that there is a certain liquidity risk in the market,and containing 5%-10% of the integrated Risk.But only La-VaR succeed in Kupiec failure rate testing, indicating the relative spread index is more reasonable liquidity risk factor, this model is effective under that index and it can capture China's stock market liquidity on a certain extent. The prediction effect of relative spread is absolutely poor, because of the high frequency data spread absolutely noisy, difficult to accurately capture the moment under the absolute spread information. ACDP model portrayed as a condition of self-regression model is not able to calculate precisely under the impact of lack of large spreads while liquidity risk was just from such a shock. Therefore, the absolute spread modeling and data frequency selected could be improved.Innovation of this paper is mainly reflected in the following three points:At first, most of the stock market liquidity indicators measure the degree day data, however this paper used a high frequency of five minutes stock spreads and returns as the research object, which can more effectively grasp the degree of liquidity and stock market information asymmetry. Secondly, this paper introduces a new liquidity factor based on an improved bid-ask spread liquidity risk model--Absolute spread index, build a model through foreign mature count time series, compare the relative spread with the original model, and finally using La-VaR results to determine the adaptability of absolute spread index in the domestic market. This action which has not been tried out before is a new exploration to calculate the liquidity risk in domestic securities market. Thirdly, few Chinese use Vine Copula method into comprehensive risk management. Therefore, innovation of this paper is using Vine Copula function to estimate multiple income dependencies and using binary Copula function to estimate spread-earnings dependencies which makes Copula to serve the La-VaR model better and offers a learning risk measure tool for for market traders.But this paper still has some shortcomings. At first, La-VaR model based on the bid-ask spread still exists limitations like considering only the width of the liquidity risk. How to use the liquidity risk factor to do multidimensional extension is the direction of future development of this article. Secondly, this article using binary Copula which is independence of Vine Copula structure to estimate the relationship between the yield and spread. In fact, this method does not fully take advantage of Vine Copula structure when used to measure dependencies between high-dimensional elements. Thus building an effective Vine Copula model containing vine returns and spreads both is a future development goal.The future research directions of this article are the following two aspects:the first is to expand liquidity indicators of La-VaR model improvements. Currently there are researchs based on the trading volume for the vine Copula method duration, we can try to combine the trading volume and bid-ask spread indicators to give a broader measure of liquidity risk. Secondly,the foreign scholars have explored a combination of returns and bid-ask spread of ten-dimensional vine Copula model,and researchs for portfolio VaR under the framework of Copulawith higher dimensions.So it's also a future development goals to build a vine Copula model include more numbers of risk factors and assets, giving full play to break the structure of the "dimension curse".